The real, current steps to opt out and request removal — plus the honest limits of what deletion can actually do.
You can ask OpenAI to stop showing your personal data in ChatGPT’s answers, and you can stop your chats from training future models. Here’s exactly how — and why, unfortunately, it’s only part of the fix.
In ChatGPT: click your profile icon → Settings → Data Controls → turn off “Improve the model for everyone.” New conversations stop being used for training immediately. For one-off privacy, use Temporary Chat, which isn’t saved or used for training.
Go to OpenAI’s Privacy Portal (privacy.openai.com), choose “Make a Privacy Request,” verify your identity, and select “Remove my personal data from ChatGPT responses.” Under laws like the GDPR and CCPA you can also request a copy or deletion of your data.
OpenAI is one company. Gemini, Copilot, Claude, Grok, and Perplexity each have their own settings and request processes. A real opt-out means doing this everywhere — and the web-connected models also pull live from data brokers, which you have to address separately.
Turning off training stops future learning. It does not remove what a trained model already absorbed.
A trained model doesn’t store you in a row an engineer can delete — your data is dissolved across billions of parameters. Europe’s “right to be forgotten” was written for databases; it can’t reach model weights. That’s why the durable answer isn’t deletion, it’s management: shrinking your exposure, and shaping the sources models retrieve so the answer they give is accurate.
Sources: OpenAI Help Center (right to be forgotten / data removal); OpenAI Privacy Portal; GDPR Article 17.
Before you file requests, see exactly what each model reveals about you. Free scan, ~60 seconds.
Scan my name →